AI use-case risk assessment
Worked example: AI-assisted AML transaction monitoring · ARA-261031-04C8B0 · 2026-10-31 06:37:15 UTC · Agentic Risk Analyst v3.2.1 · Worked example: prepared in advance, not generated live
Scenario
A Tier-2 bank in Australia proposes to replace its rules-based AML transaction-monitoring engine with an LLM-based agent that reads free-text payment memos, customer notes and adverse-media results to score suspicious-matter reports (SMRs). The agent will draft the SMR narrative and auto-lodge to AUSTRAC once a compliance analyst clicks approve. The model is hosted on a US cloud region.
Executive narrative
The idea has merit as a prioritisation and drafting aid, but replacing the rules engine without proven parity is not acceptable.
Top risks
- Suspicious matters missed after replacing rules
- Analysts approve drafts without real review
Conditions
- Do not retire the rules engine until detection parity is demonstrated
- Evidence-linked drafts and second-line sampling
Remediation plan
| Horizon | ID | Action | Owner | Priority |
|---|
| 0-30 days | A-01 | Redesign as parallel run alongside the rules engine | AML Compliance Lead | Critical |
| 30-90 days | A-02 | Build citation checks into SMR drafts | AI Engineering | High |
| 30-90 days | A-03 | Assess cross-border data handling of the US-hosted model | Privacy Officer | High |
Risk register
| ID | Risk | L | I | Inherent |
|---|
| R-01 | Suspicious matters missed after replacing rules | Possible | Severe | Critical |
| R-02 | Analysts approve drafts without real review | Likely | Major | High |
| R-03 | Customer data processed offshore | Likely | Moderate | High |
| R-04 | Hallucinated facts in SMR narratives | Possible | Major | High |
Regulatory findings
Evidence status: model knowledge only; verify against source texts.
| Regulation | Applicability | Why it matters |
|---|
| Anti-Money Laundering and Counter-Terrorism Financing Act 2006 (suspicious matter reporting) (Australia) | applicable | The agent supports SMR decisions and drafting. |
| Privacy Act 1988 (APP 8 cross-border disclosure) (Australia) | applicable | Personal information is sent to a US-hosted model. |
| APRA CPS 230 Operational Risk Management (Australia) | applicable | Material reliance on a third-party model provider for a compliance-critical process. |
Security threats
| ID | Threat | Inherent | MITRE |
|---|
| T-01 | Prompt injection through payment memos | High | OWASP LLM01:2025 |
| T-02 | Exposure of SMR content | High | - |
Controls and residual risk
| ID | Control | Priority | Addresses |
|---|
| C-01 | Parallel run with the rules engine | Immediate | R-01 |
| C-02 | Evidence-linked SMR drafts | Immediate | R-04, R-02 |
| C-03 | Quality sampling of approvals | Near-term | R-02 |
| C-04 | Data minimisation and onshore option | Near-term | R-03 |
| Risk | Inherent | Residual | Justification |
|---|
| R-01 Suspicious matters missed after replacing rules | Critical | Medium | Parallel running prevents coverage loss. |
| R-02 Analysts approve drafts without real review | High | Medium | Citations and sampling expose weak review. |
| R-03 Customer data processed offshore | High | Medium | Minimisation reduces exposure. |
| R-04 Hallucinated facts in SMR narratives | High | Low | Uncited statements are flagged. |
Decision support only, not legal or regulatory advice. AI-generated ratings and citations must be reviewed by a qualified person before they are relied on.